> Is that remotely plausible? How the hell would they know any better
> than meterologists?
I think the answer would be the "standing on the shoulders of giants" answer: they know what the weather guys know, because they tell them, plus something else. The question would be: what something else? My suggestion would be: they know how the trees are growing.
I think Roll makes it sound more mystical because he conflates things. He says: the thing that futures correlate with is temperature, so they must be predicting the temperature. Whereas a more precise explanation would be probably be: they are following an indicator (fruit growth) of something that correlates with both.
A plausible explanation might run as follows. Fruit can grow badly for lots of reasons that have to do with what happens early in the season, like too much or too little water, too hot or too cold, too late or too early a spring, etc, during a crucial formative period. It is possible that those early effects, which would have a lasting effect on fruit yield, correlate with repeating patterns produced by larger weather structures -- for example, El Nino patterns -- which differ appreciably from the heuristic simplification of random variation and simple long term averages. And the average temperature of the season might correlate with them. So that keeping track of how well the trees are growing might add a tenny bit of information to everything experts know.
That's pure speculation, of course. But I think it gives at least one answer to question "How *could* it be possible?" There could be others. We'd probably have to know scads more about large scale weather structures before we could say for sure.
But even if true, this doesn't however justify the idea that markets are the *best* way to tap the knowledge of the mass of the moderately well-informed and add it to the experts. I think it could be pretty well proven that if you wanted to do that for terrorism, you'd get scads more useful information and interpretative suggestions by subsidizing a set of listservs devoted to the topic, i.e., paying people to do the netnanny functions you do for free, and paying other experts to participate, without no obligation to talk to all comers, in order to give people who want to show off their knowledge an incentive to do so. Because what you should really care about (especially when it comes to terrorism) is not the prediction as the knoweldge it's based on -- which can be valuable even if the prevailing prediction is wrong. Markets are hugely focused on timing, in which accident play a big role. But policy analysts should more be concerned with underlying dynamic and long term trends. And they want their knowledge to be articulate, and checkable, and to come with sources attached -- none of which you get from a futures market.
In short, if you want is knowledge, vigorous public discussion will kick the market's ass every time. Except for allocation, the market only produces information as a by-product. Public discussion aims at it directly -- and produces it more efficiently.
Michael